Modeling behavioral deviations in ADLs using Inverse Reinforcement Learning

The detection of abnormalities in Activities of Daily Living (ADLs) has garnered significant attention in recent studies, with many employing deep learning techniques. This paper introduces a novel approach to analyzing ADL sequences, aimed at identifying meaningful deviations from an individual’s routine behavior. Our method offers several benefits for older adults, including timely care, early detection of health conditions to prevent deterioration, reduced monitoring burden on family members, and enhanced self-sufficiency without disrupting daily activities. We propose an Inverse Reinforcement Learning (IRL)-based method to detect behavioral abnormalities in older adults by analyzing ADL sequences. Our approach models the problem of abnormality detection in behavior sequences as a Markov Chain model. By applying the IRL method, we infer the reward function that motivates individuals to perform ADL from observed behavior trajectories. This inferred reward function is then used to identify potential behavior abnormalities through a threshold-based mechanism, where sequences with rewards below a specified threshold are flagged as potential abnormalities.